2D and 3D drawings should complement each other.
You need dimensioned orthographic drawings for making the parts, and I regard them as the more important (and easier to draw, manually or in CAD). However, isometric drawings aid visualising awkward shapes, and their real strength is the field of assembly- and explanatory- drawings.
It's common now for a small isometric "picture" to be put in some spare blank area on an otherwise orthographic drawing. Hemingway do that with the drawings and constructional notes packed with their kits.
'
There is nothing new in that of course. Very highly-skilled users can produce impressive brochure-art from CAD packages like Alibre, TurboCAD and Fusion, but Victorian engineers and architects frequently produced beautiful, tinted drawings (perhaps for the contract negotiations rather than shop or site use). Vehicle service-manuals have long used isometric cut-away and exploded images; and we of bus-pass age may recall the magnificent cut-away engineering drawings forming the The Eagle comic's centre-spread features!.
Oddly, despite the possibilities for architectural CAD, public-relations "artists' impression" drawings released by property-speculators now are often a lot rougher and less informative than in the past.
Of those CAD packages, TurboCAD (or at least the edition I bought) allows a direct choice of 2D drawing or 3D "model". Its competitors assume 3D-modelling from the start, with some means to extract from the model the elevations needed in the workshop.
'
As for "art", just examine some of the huge general-arrangement drawings produced a century or more ago, of railway-locomotives, power-station plant, etc. A wealth of repeated small parts, hidden-detail lines, cross-section hatching, etc – all starting as pencil on paper, with a loose T-square and set-squares. The draughtsmen (not many draughtswomen, at least not at design level) were also expected to be good at free-hand sketching. Even structural analysis was often carried out on the drawing-board, using techniques like Bowes' Notation (which I recall , vaguely, from A-Level) to calculate stress-vectors in frameworks.